Ask AI composer: the captured-selection chip now shares one wrapping…#271
Merged
Conversation
…+ stop (#2083) * feat(open-knowledge): make terminal launch prompt load + read + stop Tighten the docked-terminal bare-launch prompt: state the agent is in the OK desktop terminal, load the OK runtime contract, read the open file via OK MCP when one is focused, then stop — instead of the open-ended "Let's work on X" invitation. Instruction- and create-bearing launches keep the directive composers so explicit user intent is preserved. * test(open-knowledge): pin terminal-launch edge cases from review - whitespace-only instruction routes to bare launch (pins the .trim() guard) - empty createDescription still routes to directive composer (pins !== undefined) - cursor CLI added to instruction-preserved per-CLI assertion - changeset: note the bare folder/project scope reduction explicitly --------- GitOrigin-RevId: 576bc4666fe441e5e8d8776473c614a80e01cfd1
…p row (#2115) The captured-selection pill rendered on its own line above the file-context chips. Make it a sibling in the same flex-wrap row (ComposerContextChips now accepts children) so the selection and file chips share one row and only wrap on overflow; the expanded preview keeps its own line via basis-full. The pill already truncates a long label, so it no longer overflows its box. GitOrigin-RevId: f67bb2c9e72c16843db1d91613ac42d2c24b04b0
Contributor
There was a problem hiding this comment.
Automated approval from agents-private public-mirror-sync (run: https://github.com/inkeep/agents-private/actions/runs/28192255126). Source of truth is the monorepo; direct edits on inkeep/open-knowledge are overwritten on next sync.
|
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Ask AI composer: the captured-selection chip now shares one wrapping row with the file-context chips instead of stacking on its own line above them. The chips sit on the same row and only break to a second line when they overflow the available width. The selection chip continues to truncate a long file name so its label no longer spills past the chip's edge, and the expanded selection preview still drops to its own line beneath the chips.
Tighten the docked-terminal launch prompt. Opening a file, folder, or project in the desktop terminal ("Open in terminal" / file-tree + empty-space context menus) now hands the agent a minimal directive: it states the agent is running in the Open Knowledge desktop app terminal, loads the OK runtime contract, reads the open file via the OK MCP server when one is focused, then stops — instead of the previous open-ended "Let's work on X" invitation. A bare launch that isn't on a focused file (a folder right-click, empty-space, or project) deliberately reduces to "state surface, load OK, then stop" — the prior folder-scoped "Let's work on the
folder" directive is intentionally dropped in favor of that minimalism. Launches that carry explicit user intent (a typed "Open with AI" instruction, or the empty-state "Create with Claude CLI" brief) are unchanged — they keep threading that intent through. The web deep-link handoff is unaffected.